Three-dimensional junction temperature prediction method and device, electronic equipment and computer storage medium
By constructing a time-varying heat transfer equation and a thermal resistance-capacity-thermal network model for silicon carbide chips using electro-thermal-mechanical multi-field coupling, and combining it with a particle filtering algorithm, accurate prediction of the three-dimensional junction temperature of silicon carbide chips was achieved, solving the problems of inaccurate junction temperature prediction and response lag in existing technologies.
Patent Information
- Application Number
- CN202511423395.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-30
- Publication Date
- 2026-01-23
AI Technical Summary
Existing technologies cannot accurately predict the three-dimensional junction temperature of power electronic devices, especially in dynamic operating scenarios where there is response lag and error, and they cannot reflect the junction temperature distribution in three-dimensional space.
By collecting the substrate temperature and heat sink temperature of the silicon carbide chip, a time-varying heat transfer equation with multi-field coupling of electro-thermal-mechanical fields is constructed to determine the dominant heat flow path. A thermal resistance, thermal capacity, and thermal network reconstruction model is also constructed and adaptively updated using a particle filter algorithm to achieve three-dimensional junction temperature prediction.
It improves the accuracy and real-time performance of junction temperature prediction, enabling precise determination of the junction temperature of silicon carbide chips in three-dimensional structure, and adapting to nonlinear and unsteady heat conduction processes.
Smart Images
Figure CN121389436A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of thermal management technology for power electronic devices, and in particular to a three-dimensional junction temperature prediction method, apparatus, electronic device, and computer storage medium. Background Technology
[0002] Temperature is a crucial parameter affecting the lifespan and performance of power electronic devices, especially the junction temperature, which directly reflects the operating state of semiconductor components.
[0003] In existing technologies, the junction temperature of power electronic devices is typically determined using an iterative junction temperature method. This involves setting an initial junction temperature and updating it after each loss and junction temperature calculation until the junction temperature stabilizes. Once stable, the transient thermal resistance characteristics and average instantaneous loss of the power electronic device are used to calculate the instantaneous junction temperature sequence and the instantaneous maximum junction temperature. However, existing technologies assume that the heat conduction path of power electronic devices is linear and steady-state, neglecting the nonlinear thermal characteristics of the packaging material. Furthermore, the operating state of power electronic devices changes frequently, such as sudden load changes and switching frequency adjustments, leading to strong nonlinear and unsteady-state characteristics in the heat conduction process. This results in significant errors in junction temperature estimation. The iterative junction temperature method requires multiple iterations to approximate the true junction temperature, with each iteration relying on the results of the previous calculation. In the dynamic operation of power electronic devices, the junction temperature changes rapidly, while the iteration takes a certain amount of time. This results in the junction temperature estimation results failing to track the actual junction temperature changes in real time, leading to a dynamic response lag problem. Furthermore, most junction temperature iteration methods are based on one-dimensional thermal models, which can only estimate the average junction temperature at a certain point or on a certain surface of the power electronic device, and cannot reflect the junction temperature distribution in three-dimensional space.
[0004] It is evident that existing technologies are not accurate enough in predicting the junction temperature of power electronic devices and cannot determine the three-dimensional junction temperature of power electronic devices. Summary of the Invention
[0005] In view of this, it is necessary to provide a three-dimensional junction temperature prediction method to solve the problems that the existing technology is not accurate enough in predicting the junction temperature of power electronic devices and cannot determine the three-dimensional junction temperature of power electronic devices.
[0006] To address the aforementioned problems, in a first aspect, the present invention provides a three-dimensional junction temperature prediction method, comprising: The substrate temperature and heat sink temperature of the silicon carbide chip were collected, the power consumption data of the silicon carbide chip were obtained, and the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip was constructed. The dominant heat flow path in the silicon carbide chip is determined based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix. Based on the power consumption data and the time-varying heat transfer equation, a thermal resistance and thermal capacity thermal network reconstruction model of the silicon carbide chip is constructed to determine the three-dimensional junction temperature of the silicon carbide chip. In the thermal resistance and thermal capacity thermal network reconstruction model, the density of thermal resistance and thermal capacity nodes is determined based on the dominant heat flow path.
[0007] In one possible implementation, power consumption data of the silicon carbide chip is acquired, and a time-varying heat transfer equation for the electro-thermal-mechanical multi-field coupling of the silicon carbide chip is constructed, including: Obtain the real-time voltage and real-time current of the silicon carbide chip, and calculate the real-time power consumption of the silicon carbide chip based on the real-time voltage and real-time current. Based on real-time power consumption and silicon carbide chip material properties, a time-varying heat transfer equation for electro-thermal-mechanical multi-field coupling of silicon carbide chips is constructed. The time-varying heat transfer equation is as follows:
[0008] in, The material density of the silicon carbide chip is given. Let be the specific heat capacity of the silicon carbide chip, k be the thermal conductivity tensor, and P be the loss source term. For gradient operators, For temperature gradient, This represents the partial derivative of temperature with respect to time.
[0009] In one possible implementation, the dominant heat flow path in the silicon carbide chip is determined based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix, including: Calculate the temperature difference vector between the substrate temperature and the heat sink temperature, and construct the target multidimensional heat source distribution vector of the silicon carbide chip based on the temperature difference vector and the preset thermal sensitivity matrix. The target multidimensional heat source distribution vector is convolved using a preset Sobel operator to calculate the temperature gradient in each direction of the silicon carbide chip, and the heat flux density vector at each point in the silicon carbide chip is determined based on the temperature gradient. The dominant heat flow path in the silicon carbide chip is determined by a preset dynamic optimization algorithm based on the heat flux density vector. The dominant heat flow path in the silicon carbide chip is the channel where the heat flow is greater than a preset heat threshold.
[0010] In one possible implementation, the thermal resistance-thermal capacity-thermal network reconstruction model is as follows:
[0011] in, For the first j Each node t Temperature at any moment For ambient temperature, This represents the total loss of the silicon carbide chip. For the k-th order thermal resistance, Where is the thermal time constant, and N is the order, determined by the frequency of the real-time current.
[0012] In one possible implementation, a thermal resistance, thermal capacity, and thermal network reconstruction model of a silicon carbide chip is constructed based on power consumption data and time-varying heat transfer equations, including: The timing feature vector of the power consumption data of the silicon carbide chip is extracted through the timing feature extraction channel, and the temperature spatial feature vector of different nodes in the silicon carbide chip is extracted through the spatial feature extraction channel. By concatenating the temporal feature vector with the temperature space feature vector, the thermal resistance and thermal capacity parameters of the thermal resistance and thermal capacity thermal network reconstruction model are determined.
[0013] In one possible implementation, the three-dimensional junction temperature of the silicon carbide chip is determined based on a thermal resistance-thermal capacity-thermal network reconstruction model, including: Construct the state equations and observation equations of the reconstructed thermal resistance-thermal capacity-thermal network model; A pre-defined particle filter algorithm is used to adaptively update the three-dimensional junction temperature based on the state equation and the observation equation.
[0014] In one possible implementation, the state equation is:
[0015] in, , Let the random walk step size be a parameter. For process noise that follows a normal distribution, Let j be the junction temperature to be corrected in the i-th state. Let be the time-varying thermal resistance in the i-th state. Let be the time-varying heat capacity in the i-th state, and P be the heat loss source. randn() For generating random noise that follows a standard normal distribution; The observation equation is:
[0016] in, For observation output, Let P be the temperature state at the j-th temperature point, and P be the loss source. For time-varying thermal resistance, This refers to the collector-emitter voltage of the transistor. The observation noise follows a normal distribution with a mean of 0 and a variance of R.
[0017] Secondly, the present invention also provides a three-dimensional junction temperature prediction device, characterized in that it comprises: The data acquisition module is used to collect the substrate temperature and heat sink temperature of the silicon carbide chip, obtain the power consumption data of the silicon carbide chip, and construct the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip. The dominant heat flow path determination module is used to determine the dominant heat flow path in the silicon carbide chip based on the substrate temperature and the heat sink temperature combined with a preset thermal sensitivity matrix. The temperature prediction module is used to construct a thermal resistance and thermal capacity thermal network reconstruction model of the silicon carbide chip based on the power consumption data and the time-varying heat transfer equation to determine the three-dimensional junction temperature of the silicon carbide chip. The density of thermal resistance and thermal capacity nodes in the thermal resistance and thermal capacity thermal network reconstruction model is determined based on the dominant heat flow path.
[0018] Thirdly, the present invention also provides an electronic device, characterized in that it includes a memory and a processor, wherein, Memory, used to store programs; The processor, coupled to the memory, is used to execute a program stored in the memory to implement the steps in the three-dimensional junction temperature prediction method of any of the above embodiments.
[0019] Fourthly, the present invention also provides a computer-readable storage medium, characterized in that it is used to store a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the three-dimensional junction temperature prediction method of any of the above embodiments.
[0020] The beneficial effects of this invention are as follows: The three-dimensional junction temperature prediction method provided in this embodiment collects the substrate temperature and heat sink temperature of a silicon carbide chip, obtains the power consumption data of the silicon carbide chip, and constructs a time-varying heat transfer equation with electro-thermal-mechanical multi-field coupling for the silicon carbide chip. Through multi-physics field coupling, it reflects the dynamic thermal characteristics of the power device, demonstrating the nonlinear heat conduction process of the junction temperature and improving the accuracy of junction temperature prediction. By determining the dominant heat flow path in the silicon carbide chip based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix, the dominant heat flow path can be determined even when the heat transfer path is unstable, further improving the accuracy of junction temperature prediction. By constructing a thermal resistance, thermal capacity, and thermal network reconstruction model of the silicon carbide chip using power consumption data and the time-varying heat transfer equation, the energy consumption of the silicon carbide chip during operation is considered, and the three-dimensional junction temperature of the silicon carbide chip can be determined in three dimensions. This ensures both the accuracy of the silicon carbide chip junction temperature prediction and the realization of three-dimensional junction temperature prediction. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 A flowchart illustrating a three-dimensional junction temperature prediction method provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a method for determining a dominant heat flow path provided in an embodiment of the present invention; Figure 3 A flowchart illustrating a method for constructing an RC thermal network reconfiguration model according to an embodiment of the present invention; Figure 4 A flowchart illustrating a three-dimensional junction temperature adaptive update method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a three-dimensional junction temperature prediction device provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0023] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.
[0024] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.
[0025] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.
[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0027] A specific embodiment of the present invention, such as Figure 1 As shown, a three-dimensional junction temperature prediction method is disclosed, including: S101 collects the substrate temperature and heat sink temperature of the silicon carbide chip, obtains the power consumption data of the silicon carbide chip, and constructs the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip.
[0028] In this embodiment of the invention, the provided three-dimensional junction temperature prediction method is used to predict the three-dimensional junction temperature of a silicon carbide chip during operation. The silicon carbide chip refers to a power semiconductor chip made with silicon carbide as the semiconductor substrate material, including a substrate, components, and a heat sink, such as a Schottky diode or a MOSFET diode. The substrate temperature and heat sink temperature can be directly acquired by sensors, such as by measuring a thermistor. The power consumption data of the silicon carbide chip can be calculated using real-time voltage and current. It should be noted that the power consumption data should include power consumption under non-steady-state characteristics such as switching and load surges.
[0029] In this embodiment of the invention, the time-varying heat transfer equation of the silicon carbide chip, which is coupled with multiple electro-thermal-mechanical fields, reflects the influence of voltage, current, frequency, and material on temperature during the operation of the power device. This time-varying heat transfer equation, coupled with multiple electro-thermal-mechanical fields, can accurately represent the unstable heat conduction process in the silicon carbide chip, enabling accurate prediction of the junction temperature of the silicon carbide chip. The specific construction of the time-varying heat transfer equation will be described in detail later in this invention.
[0030] S102 determines the dominant heat flow path in the silicon carbide chip based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix.
[0031] In this embodiment of the invention, for most packaged silicon carbide power devices (such as MOSFETs and diodes), the dominant heat flow path is top-down: active area of the chip → silicon carbide substrate → chip mounting material → package substrate / lead frame → heat sink → final environment (usually coolant or air). This path is "dominant" because it has the lowest thermal resistance, and heat will naturally flow through the path of least resistance. However, due to frequent changes in the operating state of silicon carbide chips, such as sudden load changes and switching frequency adjustments, the heat conduction process exhibits strong nonlinear and unsteady characteristics. Consequently, the dominant heat flow path in the silicon carbide chip is not fixed, and therefore, it is necessary to determine the dominant heat flow path in the silicon carbide chip in real time. Specifically, the dominant heat flow path in the silicon carbide chip can be determined based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix. The specific determination process will be described in detail later in this invention.
[0032] S103, Based on the power consumption data and the time-varying heat transfer equation, a thermal resistance and thermal capacity network reconstruction model of the silicon carbide chip is constructed to determine the three-dimensional junction temperature of the silicon carbide chip. The density of thermal resistance and thermal capacity nodes in the thermal resistance and thermal capacity network reconstruction model is determined based on the dominant heat flow path.
[0033] In this embodiment of the invention, the thermal resistance-capacitance (RC) thermal network reconstruction model is an equivalent equation of the time-varying heat transfer equation. Thermal network reconstruction refers to the process of deriving or "reconstructing" an equivalent, simplified thermal network model with the required accuracy based on the measurement or high-fidelity simulation data of the target object (such as a complete power module). For silicon carbide chips, which have high switching speeds, high power densities, and are extremely sensitive to temperature, having an accurate thermal network model is crucial. Furthermore, the RC thermal network reconstruction model can directly reflect the three-dimensional junction temperature of the silicon carbide chip. Specifically, an RC thermal network reconstruction model of the silicon carbide chip can be constructed based on power consumption data and the time-varying heat transfer equation. The density of RC nodes in the RC thermal network reconstruction model is determined based on the dominant heat flow path. Based on the RC thermal network reconstruction model, the three-dimensional junction temperature of the silicon carbide chip can be accurately predicted.
[0034] The three-dimensional junction temperature prediction method provided in this invention collects the substrate temperature and heat sink temperature of a silicon carbide (SiC) chip, obtains the power consumption data of the SiC chip, and constructs a time-varying heat transfer equation with electro-thermal-mechanical multi-field coupling for the SiC chip. This multi-physics coupling reflects the dynamic thermal characteristics of the power device, demonstrating the nonlinear heat conduction process at the junction temperature and improving the accuracy of junction temperature prediction. By combining the substrate temperature and heat sink temperature with a preset thermal sensitivity matrix, the dominant heat flow path in the SiC chip is determined. Even when the heat transfer path is unstable, the dominant heat flow path can be identified, further improving the accuracy of junction temperature prediction. A thermal resistance, thermal capacity, and thermal network reconstruction model of the SiC chip is constructed using power consumption data and the time-varying heat transfer equation. This model considers the energy consumption of the SiC chip during operation and can determine the three-dimensional junction temperature of the SiC chip in a three-dimensional structure, ensuring both the accuracy of the SiC chip junction temperature prediction and achieving the prediction of the three-dimensional junction temperature of the SiC chip.
[0035] In some possible embodiments of the present invention, power consumption data of the silicon carbide chip is obtained, and a time-varying heat transfer equation of electro-thermal-mechanical multi-field coupling for the silicon carbide chip is constructed, including: Obtain the real-time voltage and real-time current of the silicon carbide chip, and calculate the real-time power consumption of the silicon carbide chip based on the real-time voltage and real-time current. Based on real-time power consumption and silicon carbide chip material properties, a time-varying heat transfer equation for electro-thermal-mechanical multi-field coupling of silicon carbide chips is constructed. The time-varying heat transfer equation is as follows:
[0036] in, The material density of the silicon carbide chip is given. Let be the specific heat capacity of the silicon carbide chip, k be the thermal conductivity tensor, and P be the loss source term. For gradient operators, For temperature gradient, This represents the partial derivative of temperature with respect to time.
[0037] In this embodiment of the invention, for the power consumption data of the silicon carbide chip, it is necessary to collect the real-time voltage and real-time current of the silicon carbide chip, and then calculate the real-time power consumption of the silicon carbide chip based on the power calculation formula. Therefore, the load of the silicon carbide chip may change abruptly and the working state is not fixed. Therefore, it is necessary to combine the real-time data of the silicon carbide chip to predict the junction temperature.
[0038] Furthermore, based on real-time power consumption and silicon carbide chip material characteristics, a time-varying heat transfer equation with electro-thermal-mechanical multi-field coupling is constructed for silicon carbide chips. Here, k is the thermal conductivity tensor, which is a nonlinear thermal conductivity related to temperature and mechanical stress, and P is the loss source term, which is a loss source term considering current, voltage and PWM frequency. This time-varying heat transfer equation reflects the dynamic thermal characteristics of power devices through multi-physics field coupling.
[0039] In some possible embodiments of the present invention, such as Figure 2 As shown, the dominant heat flow path in the silicon carbide chip is determined based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix, including: S201, calculate the temperature difference vector between the substrate temperature and the heat sink temperature, and construct the target multidimensional heat source distribution vector of the silicon carbide chip based on the temperature difference vector and the preset thermal sensitivity matrix. S202 uses a preset Sobel operator to perform convolution calculation on the target multidimensional heat source distribution vector, calculates the temperature gradient in each direction of the silicon carbide chip, and determines the heat flux density vector at each point in the silicon carbide chip based on the temperature gradient. S203, based on the heat flux density vector, a preset dynamic optimization algorithm is used to determine the dominant heat flux path in the silicon carbide chip, wherein the dominant heat flux path in the silicon carbide chip is the channel where the heat flow is greater than a preset heat threshold.
[0040] In this embodiment of the invention, the determination of the dominant heat flow path in the silicon carbide chip can be achieved by first collecting the temperature sensor measurements on the silicon carbide chip substrate. Radiator temperature And the total power consumption calculated from voltage and current. Calculate the temperature difference vector The target multidimensional heat source distribution vector P of the silicon carbide chip is constructed based on the preset thermal sensitivity matrix A as follows:
[0041] The Sobel operator is used to perform distributed convolution on the target multidimensional heat source distribution vector P, calculating the gradient in each direction, and then applying Fourier's law. The heat flux density vector q at each point in three-dimensional space is calculated. The calculated heat flux density vector q is visualized and analyzed. The key channels for concentrated heat flow, i.e. the dominant heat flow paths, are identified through the Dijkstra dynamic optimization algorithm, and the starting and ending points of these heat flows are determined.
[0042] Specifically, to calculate the temperature gradient and inversely infer the heat flow, the spatial gradient calculation formula for the temperature field of the IGBT chip is implemented using the Sobel operator:
[0043]
[0044] in, and Let x and y be the temperature gradients, respectively, and T be the temperature. The Dijkstra algorithm with dynamic optimization (junction node parameter R=0.1, transition node parameter C=2, and heat dissipation node parameter R=1.5) is used to find the path with the lowest thermal resistance from the heat source to the heat sink in the heat flow network, thereby achieving path optimization. The Marching Cubes algorithm is combined to realize the visualization and reconstruction of the three-dimensional junction temperature field, overcoming the limitations of the one-dimensional model.
[0045] This invention determines the accuracy of silicon carbide chip junction temperature prediction by determining the dominant heat flow path in the silicon carbide chip in real time.
[0046] In some possible embodiments of the present invention, the thermal resistance-thermal capacity-thermal network reconstruction model is as follows:
[0047] in, For the first j Each node t Temperature at any moment For ambient temperature, This represents the total loss of the silicon carbide chip. For the k-th order thermal resistance, Where is the thermal time constant, and N is the order, determined by the frequency of the real-time current.
[0048] In this embodiment of the invention, in the RC thermal network reconstruction model, thermal resistance is analogous to resistor, representing the material's ability to impede heat flow; thermal capacity is analogous to capacitor, representing the material's ability to store heat; heat flow is analogous to current, and temperature is analogous to voltage. In a chip package, a single RC pair (a thermal resistance and a thermal capacity in parallel) can represent a simple thermal mass. When power is applied, the thermal capacity begins to charge, and the temperature rises exponentially. However, the structure of actual devices (such as silicon carbide chips) is a complex stack of multiple layers of materials, each with different R and C values. Therefore, a single RC pair is far from accurate, thus requiring the use of a multi-order RC thermal network reconstruction model. The order of the RC thermal network reconstruction model can be determined by the frequency of the current, such as using order 1 for low frequencies and order 3 for high frequencies, automatically adjusting the network topology to avoid iterative lag in fixed structures. Taking a third-order RC thermal network reconstruction model as an example, specifically:
[0049] The above-described third-order RC thermal network reconfiguration model is only one possible specific embodiment of this application, and the specific form of the third-order RC thermal network reconfiguration model of this application is not limited.
[0050] In some possible embodiments of the present invention, such as Figure 3 A thermal resistance, thermal capacity, and thermal network reconstruction model for a silicon carbide chip is constructed based on power consumption data and time-varying heat transfer equations, including: S301 extracts the timing feature vector of the power consumption data of the silicon carbide chip through the timing feature extraction channel, and extracts the temperature spatial feature vector of different nodes in the silicon carbide chip through the spatial feature extraction channel. S302, the time-series feature vector and the temperature space feature vector are concatenated to determine the thermal resistance and thermal capacity parameters of the thermal resistance and thermal capacity thermal network reconstruction model.
[0051] In this embodiment of the invention, to address the problems of complex parameter mapping table construction, reliance on extensive FEA simulations, and inability to learn implicit complex features in dynamically reconstructed RC thermal network models, a spatiotemporal deep learning architecture is adopted. This architecture can directly learn and dynamically output the optimal RC thermal network parameters (R\C) from readily available spatiotemporal sequence data related to thermal behavior, replacing the traditional parameter mapping table method and achieving higher accuracy and stronger adaptive network reconstruction. For example, based on a deep learning architecture for spatiotemporal feature extraction, a dual-channel LSTM-GNN hybrid neural network is designed. The temporal channel is used to process time series of electrical parameters such as Vce(sat) and gate charge, while the spatial channel is used to process spatial distribution data such as infrared thermography and surface acoustic waves. The concept of a "thermal memory factor" is introduced to quantify the impact of historical thermal cycles on the current temperature measurement. , Here, N is the aging factor, and N is the number of thermal cycles. and The highest and average temperatures represent cumulative thermal fatigue damage; a higher value indicates a more significant impact on current thermal behavior. Specifically, the temporal feature extraction channel, composed of 2-3 layers of stacked Long Short-Term Memory (LSTM) network units, learns the temporal causal relationship between electrical parameter changes and thermal dynamic response, outputting the hidden state of the last time step as a temporal feature vector. The spatial feature extraction channel, based on a graph neural network, aggregates information from adjacent nodes, learns the spatial distribution and diffusion patterns of temperature, and performs multi-layer graph convolution operations. Global average pooling is applied to all node features to obtain a temperature spatial feature vector. The extracted temporal and temperature spatial feature vectors are concatenated and fused, mapping the fused features to a specific numerical output, which is the value of all thermal resistances R and heat capacities C in the reconstructed 3D RC thermal network at the current moment.
[0052] Furthermore, such as Figure 4 As shown, the three-dimensional junction temperature of a silicon carbide chip is determined based on a thermal resistance, thermal capacity, and thermal network reconstruction model, including: S401, construct the state equations and observation equations of the thermal resistance, thermal capacity, and thermal network reconstruction model; S402 uses a preset particle filter algorithm to adaptively update the three-dimensional junction temperature based on the state equation and the observation equation.
[0053] In this embodiment of the invention, the state equation is:
[0054] in, , Let the random walk step size be a parameter. For process noise that follows a normal distribution, Let j be the junction temperature to be corrected in the i-th state. Let be the time-varying thermal resistance in the i-th state. Let be the time-varying heat capacity in the i-th state, and P be the heat loss source. randn() For generating random noise that follows a standard normal distribution; The observation equation is:
[0055] in, For observation output, Let P be the temperature state at the j-th temperature point, and P be the loss source. For time-varying thermal resistance, This refers to the collector-emitter voltage of the transistor. The observation noise follows a normal distribution with a mean of 0 and a variance of R.
[0056] In this embodiment of the invention, when using a preset particle filtering algorithm to adaptively update the thermal resistance and thermal capacity parameters of the thermal resistance-thermal capacity thermal network reconstruction model based on the state equation and the observation equation, N particles are first generated as follows:
[0057] in, Let represent the initial particle vector of the i-th particle. This is the initial temperature range. This represents the initial thermal resistance range. This represents the initial heat capacity range.
[0058] For each particle i, the state equation is updated. , and Weights are normalized using process noise:
[0059] in, This represents the process noise of the i-th particle. exp() Represent natural numbers e An exponential function with base 1. Let be the expectation of the i-th particle. This is the inverse of the observation noise matrix.
[0060] Calculate the weights of each particle:
[0061] in, This represents the prediction noise for the i-th particle. Let N be the noise of the i-th particle, and N be the number of particles.
[0062] Specifically, 1000 particles are generated, each with a protected state variable. The RC parameters obtained from the previous embodiment are used to predict the state at the next moment and update the weight of each particle. Finally, resampling is performed to eliminate low-weight particles and replicate high-weight particles. Based on the final particle set, the optimal estimate of the system is calculated, and the three-dimensional junction temperature, shell temperature, and corrected RC parameters are finally output.
[0063] This invention breaks through the traditional single-point temperature measurement thinking, realizes three-dimensional temperature field reconstruction, quantifies the cumulative aging effect through "thermal memory factor", innovatively combines a self-destructible sensor to provide ultimate calibration, and uses a dynamic RC thermal network parameter real-time prediction method based on a dual-channel LSTM-GNN deep learning architecture, combined with a hybrid junction temperature estimation closed-loop framework of deep learning prediction and particle filter correction.
[0064] To better implement the three-dimensional junction temperature prediction method in the embodiments of the present invention, based on the three-dimensional junction temperature prediction method, correspondingly, as follows: Figure 5 As shown, this embodiment of the invention also provides a three-dimensional junction temperature prediction device, the three-dimensional junction temperature prediction device 500 comprising: The data acquisition module 501 is used to acquire the substrate temperature and heat sink temperature of the silicon carbide chip, obtain the power consumption data of the silicon carbide chip, and construct the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip. The dominant heat flow path determination module 502 is used to determine the dominant heat flow path in the silicon carbide chip based on the substrate temperature and the heat sink temperature combined with a preset thermal sensitivity matrix. Temperature prediction module 503 is used to construct a thermal resistance and thermal capacity thermal network reconstruction model of the silicon carbide chip based on the power consumption data and the time-varying heat transfer equation to determine the three-dimensional junction temperature of the silicon carbide chip. The density of thermal resistance and thermal capacity nodes in the thermal resistance and thermal capacity thermal network reconstruction model is determined based on the dominant heat flow path.
[0065] The three-dimensional junction temperature prediction device 500 provided in the above embodiments can realize the technical solutions described in the above three-dimensional junction temperature prediction method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above three-dimensional junction temperature prediction method embodiments, and will not be repeated here.
[0066] like Figure 6As shown, the present invention also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.
[0067] In some embodiments, processor 601 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in memory 602 or process data, such as the three-dimensional junction temperature prediction method of the present invention.
[0068] In some embodiments, processor 601 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 601 may be local or remote. In some embodiments, processor 601 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, or any combination thereof.
[0069] In some embodiments, memory 602 may be an internal storage unit of electronic device 600, such as a hard disk or memory of electronic device 600. In other embodiments, memory 602 may also be an external storage device of electronic device 600, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 600.
[0070] Furthermore, the memory 602 may include both internal storage units of the electronic device 600 and external storage devices. The memory 602 is used to store application software and various types of data installed on the electronic device 600.
[0071] In some embodiments, display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 603 is used to display information from electronic device 600 and to display a visual user interface. Components 601-603 of electronic device 600 communicate with each other via a system bus.
[0072] In some embodiments, when processor 601 executes the three-dimensional junction temperature prediction program in memory 602, the following steps may be performed: The substrate temperature and heat sink temperature of the silicon carbide chip were collected, the power consumption data of the silicon carbide chip were obtained, and the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip was constructed. The dominant heat flow path in the silicon carbide chip is determined based on the substrate temperature and heat sink temperature combined with a preset thermal sensitivity matrix. Based on the power consumption data and the time-varying heat transfer equation, a thermal resistance and thermal capacity thermal network reconstruction model of the silicon carbide chip is constructed to determine the three-dimensional junction temperature of the silicon carbide chip. In the thermal resistance and thermal capacity thermal network reconstruction model, the density of thermal resistance and thermal capacity nodes is determined based on the dominant heat flow path.
[0073] It should be understood that when the processor 601 executes the three-dimensional junction temperature prediction program in the memory 602, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.
[0074] Furthermore, this embodiment of the invention does not specifically limit the type of electronic device 600 mentioned. Electronic device 600 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the invention, electronic device 600 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).
[0075] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions of the three-dimensional junction temperature prediction methods provided in the above-described method embodiments.
[0076] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.
[0077] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A three-dimensional junction temperature prediction method, characterized in that, include: The substrate temperature and heat sink temperature of the silicon carbide chip are collected to obtain the power consumption data of the silicon carbide chip, and the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip is constructed. The dominant heat flow path in the silicon carbide chip is determined based on the substrate temperature and the heat sink temperature combined with a preset thermal sensitivity matrix. Based on the power consumption data and the time-varying heat transfer equation, a thermal resistance and thermal capacity network reconstruction model of the silicon carbide chip is constructed to determine the three-dimensional junction temperature of the silicon carbide chip. The density of thermal resistance and thermal capacity nodes in the thermal resistance and thermal capacity network reconstruction model is determined based on the dominant heat flow path.
2. The three-dimensional junction temperature prediction method according to claim 1, characterized in that, The process of acquiring power consumption data of the silicon carbide chip and constructing the time-varying heat transfer equation of the silicon carbide chip with electro-thermal-mechanical multi-field coupling includes: The real-time voltage and real-time current of the silicon carbide chip are obtained, and the real-time power consumption of the silicon carbide chip is calculated based on the real-time voltage and real-time current. Based on the real-time power consumption and the material properties of the silicon carbide chip, a time-varying heat transfer equation with electro-thermal-mechanical multi-field coupling is constructed for the silicon carbide chip. The time-varying heat transfer equation is as follows: in, The material density of the silicon carbide chip is given. Let be the specific heat capacity of the silicon carbide chip, k be the thermal conductivity tensor, and P be the loss source term. For gradient operators, For temperature gradient, This represents the partial derivative of temperature with respect to time.
3. The three-dimensional junction temperature prediction method according to claim 2, characterized in that, The process of determining the dominant heat flow path in the silicon carbide chip based on the substrate temperature and the heat sink temperature combined with a preset thermal sensitivity matrix includes: Calculate the temperature difference vector between the substrate temperature and the heat sink temperature, and construct the target multidimensional heat source distribution vector of the silicon carbide chip based on the temperature difference vector and the preset thermal sensitivity matrix; The target multidimensional heat source distribution vector is convolved using a preset Sobel operator to calculate the temperature gradient in each direction of the silicon carbide chip, and the heat flux density vector at each point in the silicon carbide chip is determined based on the temperature gradient. Based on the heat flux density vector, a preset dynamic optimization algorithm is used to determine the dominant heat flux path in the silicon carbide chip, wherein the dominant heat flux path in the silicon carbide chip is a channel where the heat flow is greater than a preset heat threshold.
4. The three-dimensional junction temperature prediction method according to claim 3, characterized in that, The thermal resistance, thermal capacity, and thermal network reconstruction model is as follows: in, For the first j Each node t Temperature at any moment For ambient temperature, This represents the total loss of the silicon carbide chip. For the k-th order thermal resistance, Where N is the thermal time constant and N is the order, determined by the frequency of the real-time current.
5. The three-dimensional junction temperature prediction method according to claim 4, characterized in that, The construction of the thermal resistance, thermal capacity, and thermal network reconstruction model of the silicon carbide chip based on the power consumption data and the time-varying heat transfer equation includes: The timing feature vector of the power consumption data of the silicon carbide chip is extracted through the timing feature extraction channel, and the temperature spatial feature vector of different nodes in the silicon carbide chip is extracted through the spatial feature extraction channel. The time-series feature vector is concatenated with the temperature space feature vector to determine the thermal resistance and thermal capacity parameters of the thermal resistance and thermal capacity thermal network reconstruction model.
6. The three-dimensional junction temperature prediction method according to claim 5, characterized in that, The determination of the three-dimensional junction temperature of the silicon carbide chip based on the thermal resistance, thermal capacity, and thermal network reconstruction model includes: Construct the state equations and observation equations of the thermal resistance, thermal capacity, and thermal network reconstruction model; The three-dimensional junction temperature is adaptively updated based on the state equation and the observation equation using a preset particle filtering algorithm.
7. The three-dimensional junction temperature prediction method according to claim 6, characterized in that, The state equation is: in, , Let the random walk step size be a parameter. For process noise that follows a normal distribution, Let j be the junction temperature to be corrected in the i-th state. Let be the time-varying thermal resistance in the i-th state. Let be the time-varying heat capacity in the i-th state, and P be the heat loss source. randn() For generating random noise that follows a standard normal distribution; The observation equation is: in, For observation output, Let P be the temperature state at the j-th temperature point, and P be the loss source. For time-varying thermal resistance, This refers to the collector-emitter voltage of the transistor. The observation noise follows a normal distribution with a mean of 0 and a variance of R.
8. A three-dimensional junction temperature prediction device, characterized in that, include: The data acquisition module is used to acquire the substrate temperature and heat sink temperature of the silicon carbide chip, obtain the power consumption data of the silicon carbide chip, and construct the time-varying heat transfer equation of the electro-thermal-mechanical multi-field coupling of the silicon carbide chip. A dominant heat flow path determination module is used to determine the dominant heat flow path in the silicon carbide chip based on the substrate temperature and the heat sink temperature combined with a preset thermal sensitivity matrix. The temperature prediction module is used to construct a thermal resistance and thermal capacity network reconstruction model of the silicon carbide chip based on the power consumption data and the time-varying heat transfer equation to determine the three-dimensional junction temperature of the silicon carbide chip, wherein the density of thermal resistance and thermal capacity nodes in the thermal resistance and thermal capacity network reconstruction model is determined based on the dominant heat flow path.
9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the three-dimensional junction temperature prediction method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the three-dimensional junction temperature prediction method according to any one of claims 1 to 7.